Riemannian Geometry-Preserving Variational Autoencoder for MI-BCI Data Augmentation

2026-03-11

AI Summary

This work addresses the data scarcity challenge in motor imagery brain–computer interfaces (MI-BCI) by proposing a Riemannian Geometry-Preserving Variational Autoencoder (RGP-VAE) to generate synthetic EEG covariance matrices that maintain symmetric positive-definite structure. The method uniquely integrates Riemannian geometric constraints into the generative framework through geometric mapping and a composite loss function, enabling reconstruction of covariance matrices in the tangent space while preserving their intrinsic geometric properties and establishing a subject-invariant latent space. Experimental results demonstrate that the generated covariance matrices exhibit both geometric validity and representativeness, significantly enhancing the performance of multiple classifiers on MI-BCI tasks. Furthermore, the approach supports data privacy preservation and system scalability.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

This paper addresses the challenge of generating synthetic electroencephalogram (EEG) covariance matrices for motor imagery brain-computer interface (MI-BCI) applications. Objective: We aim to develop a generative model capable of producing high-fidelity synthetic covariance matrices while preserving their symmetric positive-definite nature. Approach: We propose a Riemannian geometry-preserving variational autoencoder (RGP-VAE) integrating geometric mappings with a composite loss function combining Riemannian distance, tangent space reconstruction accuracy and generative diversity. Results: The model generates valid, representative EEG covariance matrices, while learning a subject-invariant latent space. Synthetic data proves practically useful for MI-BCI, with its impact depending on the paired classifier. Contribution: This work introduces and validates the RGP-VAE as a geometry-preserving generative model for EEG covariance matrices, highlighting its potential for signal privacy, scalability and data augmentation.

Problem

• EEG covariance matrices
• motor imagery BCI
• data augmentation
• symmetric positive-definite
• synthetic data generation

Innovation

• Riemannian geometry
• variational autoencoder
• covariance matrix
• data augmentation
• brain-computer interface